P8.15 SIMULATIONS OF THE SUPERCELL OUTBREAK OF 18 MARCH 1925

Melissa E. Becker, Matthew S. Gilmore*, Jason Naylor, Justin. K. Weber Department of Atmospheric Sciences, University of North Dakota, Grand Forks, ND

Robert A. Maddox, Tucson, AZ

Gilbert P. Compo, Jeffrey S. Whitaker, and Thomas M. Hamill NOAA Earth System Research Laboratory and Univ. of Colorado CIRES Climate Diagnostics Center, Boulder, CO

1. INTRODUCTION*

On 18 March 1925, the "Tri-State Tornado" left a path of nearly complete destruction across portions of Missouri, Illinois, and Indiana and killed almost 700 people - the deadliest tornado in U.S. history (Changnon and Semonin 1966). The Weather Research and Forecasting Model (WRF) is being used to conduct storm-scale simulations of this event in an effort to understand the environmental conditions and environmental changes that led to the long-track tornadic supercell thunderstorm and overall severe weather outbreak (Fig. 1).

2. EVENT OVERVIEW*

The Tri-State Tornado was first reported shortly after 1 PM CST north of Ellington, MO (Fig. 2). Around 2 PM, the town of Biehle, MO was struck before the storm crossed the Mississippi River and destroyed the town of Gorham, IL a half hour later. Four other towns in southern Illinois were hit by the tornado within the Fig. 1. Summary of severe weather reports from 18 March hour. By 3 PM, the storm had crossed into Indiana, 1925 (from Maddox et al. 2011) within the outbreak region continuing to destroy towns and farms in its path before (outlined in red). Several tornadoes (tracks shown), finally lifting around 3:30 PM (Changnon and Semonin thunderstorms (“ ” symbols), and widespread hail (“H” symbols) 1966). were reported. The area encompassing the 3-4 April 1974 outbreak is also overlaid for reference (purple dashed). Maddox et al. (2011) have carefully plotted and analyzed hourly synoptic maps and found that the Tri- State Tornado and other storms that day were spawned from a low pressure system that moved out of the lee of the Colorado Rockies and into northeastern Oklahoma by 13 UTC (Fig. 3). The low pressure system brought rain to areas of Missouri and Illinois overnight and into the morning hours of the 18th (hatched region of Fig. 3), before the severe storms of interest occurred, creating a pool of cooler air that slowed the northward progression of the warm front (Maddox et al. 2011). The supercell that spawned the Tri-State Tornado appeared to form close the triple point and followed the warm frontal baroclinic zone for the majority of its track (Maddox et al. 2011).

* Corresponding Author Address: Dr. Matthew Gilmore, Dept. of Atmos. Sciences, 4149 University Avenue Stop 9006, Grand Forks, ND 58202-9006. Fig. 2. The reported Tri-State Tornado 219 mi damage path (as [email protected] analyzed by Changnon and Semonin 1966).

1

3. METHODOLOGY

3.1. WRF Model Initial Conditions

The WRF simulations were initialized with 3-D atmospheric conditions from the 20th Century Reanalysis Project. Using only surface pressure as input, the reanalysis uses NCEPʼs operational global Climate Forecast System model to create an ensemble of possible 3-D atmospheric states (Compo et al. 2006, 2009; Whitaker et al. 2004): particularly useful in years predating rawinsonde observations such as 1925. The version of the NCEP ensemble model output used herein was that which resulted in 56 separate ensemble members Fig. 4. Correlation results of top three members. with 2.5° horizontal resolution and 25 vertical levels.

3.2. WRF Model Grid and Physics

Simulations were performed on NCSAʼs Intel 64 Cluster (Abe) with version 3.2 of the WRF model. Nests 1, 2 and 3 were initialized at 00 UTC 18 March 1925 with reanalysis data from Member B. Nests 4 and 5 were initialized at 12 UTC 18 March 1925 (Fig. 5). WRF also uses reanalysis data at 18 UTC 18 March 1925 but only as boundary conditions on the outer nest.

Fig. 3. Overview of the synoptic pattern at 13 UTC on 18 March 1925 as reanalyzed by the US Weather Bureau. Note the large area of rain (hatched area) north of the warm front.

Sea level pressure observations over the Tri-State region1, many of which were not originally used as input to the climate model, were objectively compared to sea level pressure of each ensemble member from the climate model using the Model Evaluation Tools (MET) software package from NCARʼs Developmental Testbed Center (DTC). This enabled choosing the ensemble member with the best 3-D state. Three of the 56 ensemble members with the highest correlations at Fig. 5. The five nests used in the WRF runs. three separate times (0, 12, and 18 UTC) are shown (Fig. 4). These nests, physics packages (Table 1), and the Member B was chosen out of these three for the model parameters are similar to that of Shafer et al. WRF simulation because not only was the correlation (2009). The outermost nest, nest one, has a dimension stronger at 12 and 18 UTC, closer to the time of the of 50X30 and has grid point spacing of 162 km. Nest severe weather outbreak, but it also had the overall best two, which covers most of the United States and part of correlation values. Furthermore, 18 UTC is a time Mexico and the Caribbean, has 54 km grid spacing and where the climate model did not assimilated the detailed dimensions of 94X58. The third nest focuses on the sea level pressures over the Tri-State area. Thus, that Midwest, the Southeast, and East Coast with time should be the best independent check of an dimensions of 178X124 and a grid spacing of 18 km. ensemble memberʼs performance. Member B Nest four (used to create Figs. 6 and 7) has dimensions performed best at 18 UTC. of 304X202 and 6 km grid spacing. The innermost nest, nest five, concentrates on the Tri-State area with 2 km 1 Observational surface data was digitized from handwritten grid spacing and dimensions of 637X466. U.S. Weather Bureau forms that indicate standard station observations, but with dewpoint only recorded three times a day (01 and 13 UTC and local noon).

2

Table 1. Physics selections within version 3.2 WRF model 4.RESULTS Physics Selection Microphysics WDM-6 (2 moment) Comparisons between the subjective analysis and WRF model output at 13 and 18 UTC show similarities in the Planetary Boundary Layer YSU location of the low although the model low weakens between those two times and has more elongation in Surface Layer Noah land-surface shape (Fig. 6). The surface temperature and dewpoint temperature contours are similar between the Convective Kain-Fritsch observations and WRF simulation except that the dewpoint temperatures are too cool at 13 UTC (Fig. 6). Radiation Dudhia (shortwave), The dewpoint in WRF, however, seems to recover by 18 rrtm (longwave) UTC in the Tri-State area.

Observations WRF Model Simulations

13 UTC

18 UTC

Fig. 6. Comparison of subjective analysis (left panels) and WRF model output (right panels) for 13 UTC (top) and 18 UTC (bottom). Contoured in black is pressure (every 4 hPa), in red is temperature (50 and 70°F shown only), and in green is dewpoint (40 and 60°F shown only).

3

Fig.7. Sequence of simulated radar reflectivity images from 18 UTC to 21 UTC on nest #5 (2 km horizontal gridspacing).

Another difference shown in Fig. 6 is that the 7b and 7c) which is about one hour prior to that of the outflow boundary/warm front that was distinct from the actual Tri-State supercell. Then the simulated separate pressure trough to the NE of the low is not dominant storm crosses into Indiana at about 21:15 distinct in the model simulation. More investigation is (just after Fig. 7d), about 45 minutes early, but at the needed to see whether the overnight rainfall was same location as the actual Tri-State supercell. pronounced to the NE of the low. Thus, even though the simulated dominant storm Remarkably, there is a dominant cell in the line occurs too early, its motion and path is strikingly that originates South Central Missouri between 17 similar to the actual Tri-State supercell. Also, and 18 UTC (visible in simulated radar reflectivity at although the dominant storm seems to take on bow- 18 UTC - Fig 7a). The cell crosses the border from echo characteristics in these plots, more work is Missouri into Illinois at about 19:30 UTC (between Fig needed to investigate its rotational properties.

4

Also, it should be mentioned that some cells in shown) near where some of the other tornadic later reflectivity images in the warm sector take on a supercells occurred on this day (Fig. 1). More work is kidney-bean shape reminiscent of supercells (not needed to investigate the cell behavior in those cases with re-simulations at finer horizontal grid spacing. What were the environmental parameters supporting these simulated storms? Interestingly, the CAPE values are quite small in the vicinity of the low pressure system (500 to 1000 J kg–1 or less in Fig. 8) - consistent with cool-season supercells (Johns et. al. * 1993). Since the dewpoint temperature and temperature fields are in good agreement (simulation versus observations – Fig. 6) at 18 UTC, and since the boundary layer moisture is quite deep (Fig. 11), perhaps the Tri-State Tornado outbreak was indeed similar to other cool-season tornado outbreaks. However, it is also possible that CAPE might be erroneously small due to warm midlevel temperatures

Fig. 8 Most unstable Convective Available Potential Energy (500-300 hPA) in the reanalysis. (CAPE) at 18 UTC with warmer colors corresponding to The vertical wind shear values at 18 UTC are larger values (see key). The asterisk denotes the largest along the baroclinic zone (Fig. 9), as expected, approximate location of Baker, MO. and are consistent with that which would support supercells (over 25 m/s of shear in the lowest 6 km). The 0-3 km SRH values within most of the warm sector are greater than 250 m2 s–2 (Fig. 10): the mean value associated with F4 tornadoes (Kerr and Darkow 1996). Extreme values of 0-3 km SRH > 500 are found immediately along the warm front. (The small values of 0-6 km wind shear and 0-3 km SRH along the warm front in Figs. 9-10 are associated with regions where storms have perturbed the environmental winds.)

5. SUMMARY AND FUTURE WORK

The Tri-State tornado of 1925 remains the deadliest tornado in U.S. history. To better understand the atmospheric conditions that were in Fig. 9. Vertical wind shear from 0 to 6 km at 18 UTC with place to support this long-lived tornado, the WRF warmer colors corresponding to larger values (see key). model was initialized with reanalyses from the Climate Forecast Systems Model. After choosing the ensemble member with output best correlated to actual sea level pressure observations and running WRF, the output of sea level pressure, temperature, and dewpoint temperature were compared to subjective analysis. A simulated dominant cell moved in a similar direction and speed as the Tri-State supercell. The output of CAPE, shear, and helicity suggest that the tornadic supercell may have formed in an area of high shear and low CAPE, consistent with cool-season events. Future work will incorporate independent temperature and dewpoint readings in addition to sea level pressure when choosing the best ensemble member, testing sensitivity to different parameterization schemes, and tornado-resolving supercell simulations. Also, higher resolution Fig. 10. Storm-relative helicity (0-3 km) at 18 UTC with reanalyses may be used as WRF input. warmer colors corresponding to larger values (see key).

5

Compo, G. P. et. al, 2009: The Twentieth Century Reanalysis Project. Quart. J. Roy. Met. Soc., in preparation.

Compo,G. P., J. S. Whitaker, and P. D. Sardeshmukh, 2006: Feasibility of a 100 year reanalysis using only surface pressure data. Bull. Amer. Met. Soc., 87, 175-190.

Johns, R. H., J. M. Davies, and P. W. Leftwich, 1993: Some wind and instability parameters associated with strong and violent tornadoes. Part 2. Variations in the combinations of wind and instability parameters. The Tornado: Its Structure, Dynamics, Prediction, and Hazards, Geophys. Monogr., No. 79, Amer. Geophys. Union, 583-590.

Kerr, B. W., and G. L. Darkow, 1996: Storm-relative winds and helicity in the tornadic thunderstorm environment. Wea. Forecasting, 11, 489–505. Fig. 11. Sounding from Baker, MO, 19 UTC. The most unstable CAPE on this sounding is approximately 700 J kg–1. Maddox, R. et al., 2011: Meteorological analyses of See Fig. 8 for the approximate location of Baker,MO. the Tri-State Tornado event of 18 March 1925. To be submitted to Mon. Wea. Rev.

6. ACKNOWLEDGEMENTS Shafer, C. M.; A. E. Mercer, C. A. Doswell, M. B.

Richman, L. M. Leslie, 2009: Evaluation of Computer time was provided by Teragrid grant TG- WRF forecasts of tornadic and non-tornadic ATM100039. This research was supported by NSF outbreaks when initialized with synoptic-scale grant AGS-0843269. The Intercollegiate Academic input Mon. Wea. Rev.,137, 1250 Fund provided partial travel expenses for Ms. Becker.

Whitaker, J.S., G.P.Compo, X. Wei, and T.M. Hamill

2004: Reanalysis without radiosondes using 7. REFERENCES ensemble data assimilation. Mon. Wea. Rev.,

132, 1190-1200. Changnon, S. A., and R. G. Semonin, 1966: A great tornado disaster in retrospect. Weatherwise, 19, 56-65.

6